Search for: All records

This study presents an initial model for bark beetle identification, serving as a foundational step toward developing a fully functional and practical identification tool. Bark beetles are known for extensive damage to forests globally, as well as for uniform and homoplastic morphology which poses identification challenges. Utilizing a MaxViT-based deep learning backbone which utilizes local and global attention to classify bark beetles down to the genus level from images containing multiple beetles. The methodology involves a process of image collection, preparation, and model training, leveraging pre-classified beetle species to ensure accuracy and reliability. The model's F1 score estimates of 0.99 and 1.0 indicates a strong ability to accurately classify genera in the collected data, including those previously unknown to the model. This makes it a valuable first step towards building a tool for applications in forest management and ecological research. While the current model distinguishes among 12 genera, further refinement and additional data will be necessary to achieve reliable species-level identification, which is particularly important for detecting new invasive species. Despite the controlled conditions of image collection and potential challenges in real-world application, this study provides the first model capable of identifying the bark beetle genera, and by far the largest training set of images for any comparable insect group. We also designed a function that reports if a species appears to be unknown. Further research is suggested to enhance the model's generalization capabilities and scalability, emphasizing the integration of advanced machine learning techniques for improved species classification and the detection of invasive or undescribed species. 

Abstract Tropical elevation gradients support highly diverse assemblages, but competing hypotheses suggest either peak species richness in lowland rainforests or at mid‐elevations. We investigated scolytine beetles—phloem, ambrosia and seed‐feeding beetles—along a tropical elevational gradient in Papua New Guinea.Highly standardised sampling from 200 to 3700 m above sea level (asl) identified areas of highest and lowest species richness, abundance and other biodiversity variables.Using passive flight intercept traps at eight elevations from 200 to 3500 m asl, we collected over 9600 specimens representing 215 species. Despite extensive sampling, species accumulation curves suggest that diversity was not fully exhausted.Scolytine species richness followed a unimodal distribution, peaking between 700 and 1200 m asl, supporting prior findings of highest diversity at low‐to‐mid elevations.Alternative models, such as a monotonous decrease from lowlands to higher elevations and a mid‐elevation maximum, showed lesser fit to our data. Abundance is greatest at the lowest sites, driven by a few extremely abundant species. The turnover rate—beta diversity between elevation steps—is greatest between the highest elevations.Among dominant tribes—Dryocoetini, Xyleborini and Cryphalini—species richness peaked between 700 and 2200 m asl. Taxon‐specific analyses revealed distinct patterns:Euwallaceaspp. abundance uniformly declined with elevation, while other genera were driven by dominant species at different elevations.Coccotrypesand phloem‐feedingCryphalushave undergone evolutionary radiations in New Guinea, with many species still undescribed. Species not yet known to science are most likely to be found at lower and middle elevations, where overall diversity is highest. 

September 2023 marked the 50th anniversary of the Electronic Visualization Laboratory (EVL). This paper summarizes EVL’s efforts in Visual Data Science, with a focus on the many networked, immersive, collaborative visualization and virtual-reality (VR) systems and applications the Lab has developed and deployed, as well as lessons learned and future plans. 

We develop FLM, a high-level language that enables network operators to write programs that recognize and react to specific packet sequences. To be able to examine every packet, our compilation procedure can transform FLM programs into P4 code that can run on programmable switch ASICs. It first splits FLM programs into a state management component and a classical regular expression, then generates an efficient implementation of the regular expression using SMT-based program synthesis. Our experiments find that FLM can express 15 sequence monitoring tasks drawn from prior literature. Our compiler can convert all of these programs to run on switch hardware in way that fit within available pipeline stages and consume less than 15% additional header fields and instruction words when run alongside switch programs. 

Objective: Persistent Identifiers (PIDs) are central to the vision of open science described in the FAIR Principles. However, the use of PIDs for scientific instruments and facilities is decentralized and fragmented. This project aims to develop community-based standards, guidelines, and best practices for how and why PIDs can be assigned to facilities and instruments. Methods: We hosted several online and in-person focus groups and discussions, cumulating in a two-day in-person workshop featuring stakeholders from a variety of organizations and disciplines, such as instrument and facilities operators, PID infrastructure providers, researchers who use instruments and facilities, journal publishers, university administrators, federal funding agencies, and information and data professionals. Results: Our first-year efforts resulted in four main areas of interest: developing a better understanding of the current PID ecosystem; clarifying how and when PIDs could be assigned to scientific instruments and facilities; challenges and barriers involved with assigning PIDs; incentives for researchers, facility managers, and other stakeholders to encourage the use of PIDs. Conclusions: The potential for PIDs to facilitate the discovery, connection, and attribution of research instruments and facilities indicates an obvious value in their use. The lack of standards of how and when they are created, assigned, updated, and used is a major barrier to their widespread use. Data and information professionals can work to create relationships with stakeholders, provide relevant education and outreach activities, and integrate PIDs for instruments and facilities into their data curation and publication workflows. 

Per- and polyfuoroalkyl substances (PFAS) in the environment pose persistent and complex threats to human and wildlife health. Around the world, PFAS point sources such as military bases expose thousands of populations of wildlife and game species, with potentially far-reaching implications for population and ecosystem health. But few studies shed light on the extent to which PFAS permeate food webs, particularly ecologically and taxonomically diverse communities of primary and secondary consumers. Here we conducted >2000 assays to measure tissue-concentrations of 17 PFAS in 23 species of mammals and migratory birds at Holloman Air Force Base (AFB), New Mexico, USA, where wastewater catchment lakes form biodiverse oases. PFAS concentrations were among the highest reported in animal tissues, and high levels have persisted for at least three decades. Twenty of 23 species sampled at Holloman AFB were heavily contaminated, representing middle trophic levels and wetland to desert microhabitats, implicating pathways for PFAS uptake: ingestion of surface water, sediments, and soil; foraging on aquatic invertebrates and plants; and preying upon birds or mammals. The hazardous long carbon-chain form, perfuorooctanosulfonic acid (PFOS), was most abundant, with liver concentrations averaging >10,000 ng/g wet weight (ww) in birds and mammals, respectively, and reaching as high 97,000 ng/g ww in a 1994 specimen. Perfuorohexanesulfonic acid (PFHxS) averaged thousands of ng/g ww in the livers of aquatic birds and littoral-zone house mice, but one order of magnitude lower in the livers of upland desert rodent species. Piscivores and upland desert songbirds were relatively uncontaminated. At control sites, PFAS levels were strikingly lower on average and different in composition. In sum, legacy PFAS at this desert oasis have permeated local aquatic and terrestrial food webs across decades, severely contaminating populations of resident and migrant animals, and exposing people via game meat consumption and outdoor recreation. 

Lake sturgeon (Acipenser fulvescens) is a species of conservation concern that has been stocked in several Great Lakes (North America) rivers. Lake sturgeon were extirpated in the Ontonagon River in Lake Superior and stocking began in 1998. In 2017, gametes were collected from spawning lake sturgeon (9 females, 36 males) caught at the nearby Sturgeon River spawning ground, generating nine family groups using a 1:4 mating design (n = 862). In 2018, gametes were collected from 3 females and 15 males, generating three family groups, and additional collections of drifting fry from the Sturgeon River were reared in the hatchery, resulting in 84 hatchery-produced and 675 wild-caught fry for stocking in the Ontonagon River. The objective of this study was to compare paternal representation and genetic diversity between the two stocking strategies. Parentage analysis based on genetic data from 12 microsatellite loci determined none of the family groups in the hatchery had equal paternal representation (p < 0.001), while wild-produced offspring had equal paternal representation. Despite the larger number of breeders contributing to the wild-caught larvae, there was no significant difference in genetic diversity between the wild-caught larvae and representative hatchery-produced offspring.